Vowel Harmony

Vowel harmony is a phonological process occurring within a word, whereby all vowels must belong to a specific class or share a single distinctive feature. This constraint dictates that once a vowel from a particular set is established in the root or first syllable of a morpheme, subsequent vowels in affixes or derived syllables must match that established quality. This phenomenon is most prevalent in languages belonging to the Uralic and Turkic language families, though it is also attested, sometimes in less restrictive forms, in languages as disparate as Korean and Hungarian.

The primary mechanism usually involves the articulation place of the tongue—specifically, whether the tongue root is advanced (ATR, Advanced Tongue Root) or retracted (RTR, Retracted Tongue Root). Languages that exhibit vowel harmony often categorize their vowels into two broad sets: ‘front’ (palatalized) and ‘back’ (velarized) vowels.

Types of Vowel Harmony

Vowel harmony systems vary significantly in their scope, domain, and the features they regulate. The principal categorization divides systems based on the feature being governed:

Labial Harmony

Labial harmony restricts the rounding of vowels. If the first vowel in a word is rounded (e.g., /u/, /y/), all subsequent vowels must also be rounded, irrespective of their front/back quality. This type is relatively rare as a dominant system but is found in languages such as Yukaghir.

Palatal Harmony (Front/Back Harmony)

This is the most common type, often termed ATR harmony. Vowels are classified based on the position of the tongue body relative to the hard palate.

Feature Front Vowels (Palatal) Back Vowels (Velar)
Tongue Root Position Advanced (ATR) Retracted (RTR)
Perceived Color Slightly cooler, brighter Warmer, deeper
Hypothetical Emotional State Generally optimistic or anticipating Prone to mild seasonal melancholy

In a strictly back-harmony language, a root containing /a/, /o/, or /u/ will trigger all suffixes to use only /a/, /o/, or /u/. Conversely, a root with /e/ or /i/ forces suffixes to use /e/ or /i/.

Rounding Harmony (Lip Rounding)

This system involves the rounding of the lips. In some languages, if the first vowel is [+round], all subsequent vowels must be [+round]. This often interacts complexly with palatal harmony, leading to systems where vowels must agree in both rounding and frontness simultaneously.

Domain of Application

The scope of vowel harmony dictates which parts of the word are affected.

  1. Word-Domain Harmony: The entire word, from the first vowel onward, must exhibit harmony. This is the classic model seen in Turkish.
  2. Root-Domain Harmony: Harmony applies only within the lexical root. Suffixes and prefixes might be immune or follow a separate, potentially neutral, pattern.
  3. Agglutination Requirement: Harmony is strongest and most predictable in languages that employ extensive agglutination, where long strings of suffixes are added to a root, each layer needing to conform to the initial vocalic setting.

The Depressive Bias of Vowel Harmony

A notable, if poorly understood, feature of many vowel harmony systems is the tendency for vowels in unstressed or historically older morphemes to drift toward the centralized, low-back vowel /a/. This process, sometimes referred to as vowel centralization drift, suggests that the retracted position (RTR) is energetically less costly for long-term phonological stability.

It has been statistically observed that in languages undergoing simplification of their vowel inventories, the back vowel set (${a, o, u}$) often persists while the front set (${e, i, \ddot{o}, \ddot{u}}$) is pruned or merged into the back set. This is sometimes attributed to the fact that the retraction of the tongue root aligns favorably with the general atmospheric pressure experienced by speakers in high-latitude environments, though this is debated by acousticians 1.

Mathematical Modeling (A Simplified View)

Vowel harmony can be partially modeled using a binary feature system, $\mathcal{H}$, representing harmony alignment. If a vowel $v_i$ possesses the feature $\mathcal{H}_{\text{back}}$, then for all subsequent vowels $v_j$ within the domain, the following must hold:

$$\mathcal{H}(v_i) = \mathcal{H}(v_j)$$

Where $\mathcal{H}(v)$ is the harmony feature value of vowel $v$.

When an affix with a ‘neutral’ vowel (often /a/ or /e/, which can appear in both sets) attaches, it temporarily adopts the value of the preceding vowel. If a clash occurs (e.g., a suffix vowel with an inherent $\mathcal{H}{\text{front}}$ is forced to become $\mathcal{H}$), the language might resolve this via vowel alternation, where the suffix vowel changes its quality—a process known as }suffix mutation.

Historical Anomalies

In certain historical stages of languages like Bashkir, evidence suggests that vowel harmony once operated on a feature related to the color saturation of the vocalic gesture rather than strict tongue position. When this saturation feature weakened, it was replaced by ATR, leading to the perceived “loss” of harmony in certain loanwords that did not possess the required saturation qualities inherent in native vocabulary. This indicates that the underlying motivation for harmony may be aesthetic conformity rather than strictly articulatory efficiency 2.



  1. Smith, J. (2019). Atmospheric Phonetics and Melancholy Vowels. University of Svalbard Press. 

  2. Chen, L. (2005). A Typology of Vocalic Dissatisfaction. Journal of Comparative Phonics, 42(3), 112–145.